I�ve just begun to study these things [*], so maybe I
can say something about this subject

Dan Minette wrote:
> 
> The reason this is important is that it is tied into
> the proof of true statements in a formal system.
> My understanding of what Godel proved is
> "every formal system at least as complicated as arithmetic
> will have true statements that cannot be proven within
> the system."  These statements are
> (usually always?) self-referential statements.
>
The math that studies "proof" is called Model Theory, and
it goes something like this: in math, we have a structure
[for example, a group with one "special" element and a
binary operation] and a set of axioms that this structure
must satisfy [again, in a group we have the neutral element,
the inverse, and associativity]. In Model Theory, we have
a "language" [that can even have an infinite number of
symbols, that generates sentences], and the set of axioms
[again, it can be infinite], and proof of truth or falsity
comes from rules that are clearly defined for manipulating
sentences.

Truth or Falsehood of a sentence are defined if for
all structures that satisfy the axioms the sentence is
valid in that structure.

It�s possible to say that some sentences are "false"
or that some statements are "true" [examples of both:
"exists x such that x != x" and "for all x, x = x"],
while some other sentences may or may not be valid
["exists x, exists y, such that x != y" is valid in
any structure with 2 or more elements, false otherwise].

One curious outcome of Model Theory is that there is
_no_ system of axioms that can prove that the
structure has a finite number of elements without
fixing this number of elements [an axiom that
says that the structure has at least 3 elements is
"exists x, exists y, exists z, such that x != y and
x != z and y != z", while and axiom that says
that the structure has at most 3 elements is
"for all x, for all y, for all z, for all t,
x = y or x = z or x = t or y = z or y = t or z = t";
take both axioms together and we have a new theorem
equivalent to saying that there are 3 and only 3
distinct elements]

BTW, these two axioms that define a set with 3 elements
are "complete", because they define _uniquely_ [up
to isomorphism] a set of three elements.

Godel�s theorem is that any system
of axioms [this includes an _enumerable_ system of
axioms, but I don�t know if more than that are allowed]
that is complex enough to include set theory is not
"complete", in other words, there are structures
that satisfy them and that are _not_ isomorphic.

For example, if we take ZFC [the Zermelo-Frankel Axioms
of Set Theory with the Continuum Hypothesis] it�s even
possible to _construct_ an _enumerable model_ of it.
Namely, it�s possible to build a set U and a binary 
relation @ [damn keyboard without math symbols] such
that (U,@) satisfy _all_ ZFC axioms, with @ playing the
role of the "belongs to" relation!!! 

[and it�s even intuitive how to build this model: just
take as 0-th step the empty set, and as n-th step all
the sets that can be build from the previous set using
(once) all axioms; each step is enumerable, so its union
is enumerable too]

> Humans have means of handling these statements.
> Humans can find ways to prove these statements
> outside of the formal system. This is one of the main
> arguments against considering human thinking as
> algorithmic: we can handle these self-referential
> statements.
>
But humans _can�t_ handle these statements, heck,
there are even statements that are defined in a _finite_
set that humans can�t handle. What is the most significant
digit of 2^(10^googol) ? 

No matter how smart you think a human brain might be,
we are _limited_ in time to run an algorithm.

> Its not that humans can make such predictions; they can't.
> Its that the ability to handle what humans do
> algorithmically involves the completeness
> of algorithms.  That's why the question of a universal
> Turing machine is important.
>
Is it really important? We know that there are _finite_
problems that can�t be solved by machine or human brain.
So why should we be so scared that there are _infinite_
problems that can�t be solved?

Alberto Monteiro

Reply via email to